Hybrid effort estimation based on multivariate liner regression and analogy based estimation

We propose a hybrid effort estimation method based on the multivariate liner regression analysis and the analogy based estimation method (ABE). First, our method calculates the unreliability index of ABE on an estimation target project. Next, our method selects log-log regression estimation when the value of the index is low. Otherwise our method selects ABE estimation or combined estimation (the average of ABE and log-log regression estimation). In the experiment, we compared estimation accuracies of our method with conventional methods, and the results showed that the median of Balanced Relative Error (estimation accuracy index) was improved from 47.2% to 39.7%, when the variance of similar projects' effort was used as the reliability index.